Low probability of activity limitation is one criterion for successful ageing( Reference Rowe and Kahn 1 ). Activity limitation results from a long-standing process, known as the ‘disablement process’( Reference Verbrugge and Jette 2 ). Nutrition may influence this dynamic process( Reference Inzitari, Doets and Bartali 3 ). Higher intake of some individual foods, particularly vegetables, fruit, dairy products and some micronutrients, were associated with a reduced risk of activity limitation( Reference Vercambre, Boutron-Ruault and Ritchie 4 – Reference Sharkey, Branch and Giuliani 8 ). However, as nutrients may interact with others, studying them in combination in dietary patterns has been encouraged when analysing the role of diet in the disablement process( Reference Inzitari, Doets and Bartali 3 ). Higher adherence to a Mediterranean diet – that is predefined healthy dietary pattern – was associated with a slower decline of mobility and a lower risk of developing new mobility limitation over 9 years( Reference Milaneschi, Bandinelli and Corsi 9 ), and a lower risk of limitation in activities of daily living in older women over 5 years( Reference Féart, Pérès and Samieri 10 ). A healthy diet assessed by the Healthy Eating Index score was also associated with a reduced risk of activity limitation( Reference Koster, Penninx and Newman 11 , Reference Xu, Houston and Locher 12 ). Thus far, only one Japanese study analysed the relationship between a posteriori dietary patterns – that is derived independently of any prior hypotheses about their beneficial effect on health – and the 5-year risk of activity limitation in adults aged 73·9 (sd 6·0) years on average( Reference Tomata, Watanabe and Sugawara 13 ). In this study, the Japanese diet, characterised by high intake of fish, vegetables, mushrooms, potatoes, seaweeds, pickles, soyabean and fruit, was associated with a lower risk of incident activity limitation. As a posteriori patterns are population- and cultural-specific, extrapolating these data to other settings is difficult. In a previous study, we derived five sex-specific dietary clusters in older French community-dwellers from the Three-City (3C) Bordeaux study( Reference Samieri, Jutand and Féart 14 ). Specific unhealthy clusters were associated with worse cognitive function, and higher depressive symptomatology( Reference Samieri, Jutand and Féart 14 ), both being risk factors for activity limitation( Reference Gill, Williams and Richardson 15 , Reference Schillerstrom, Royall and Palmer 16 ). We also showed that some dietary clusters were associated with incident frailty( Reference Pilleron, Ajana and Jutand 17 ), which is a predictor of disability but is not equivalent to it( Reference Fried, Tangen and Walston 18 ).
We therefore investigated the association between dietary clusters and the 10-year risk of activity limitation in French older adults in the Bordeaux sample of the 3C Study.
Methods
Study population
The study was conducted in the Bordeaux sample of the 3C Study, a prospective cohort study of vascular risk factors for dementia that included 2104 community-dwellers aged 65 years or older in 1999–2000( 19 ). The protocol and baseline characteristics have been detailed previously( 19 ). In brief, all individuals aged 65 years and older living in Bordeaux, France, registered on the electoral rolls and not institutionalised were eligible. The participants’ selection was carried out in each administrative district, with individuals sampled randomly proportional to the district’s population( 19 ). At baseline, the standard data collection performed by trained psychologists included socio-demographic and lifestyle characteristics, medical history, neuropsychological testing, a physical examination and blood sampling. Five follow-up examinations were performed 2 years (wave 1, 2001–2002), 4 years (wave 2, 2003–2004), 7 years (wave 3, 2006–2007), 10 years (wave 4, 2009–2010) and 12 years (wave 5, 2011-2012) later.
A 24-h dietary recall and a detailed food frequency questionnaire were administered by trained dietitians at wave 1; this analysis was therefore based on data from wave 1 (hereinafter referred to as baseline) to wave 5.
Fig. 1 portrays the detailed flow chart of analysis samples. Briefly, out of 1755 participants at baseline, we retained 583 (274 men and 309 women) for the analysis of the risk of mobility restriction, 1114 (451 men and 663 women) for the analysis of the risk of instrumental activities in daily living (IADL) and 1267 (491 men and 776 women) for the analysis of the risk of activities in daily living (ADL).
Flow chart of studied samples, The Three-City Bordeaux Study. FU, follow-up; ADL, activities in daily living; IADL, instrumental activities in daily living.
Ethical considerations
This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human subjects were approved by the Consultative Committee for the Protection of Persons participating in Biomedical Research of the Kremlin-Bicêtre University Hospital (Paris). Written informed consent was obtained from all subjects.
Dietary assessment
A FFQ and a 24-h dietary recall were administered by specifically trained dietitians during the same interview( Reference Féart, Jutand and Larrieu 20 ). For the 24-h dietary recall, portions size was assessed using photographs. The number of weekly servings of food obtained from the FFQ had moderate but significant correlations (0·20–0·43), with nutrient intakes obtained from the 24-h dietary recall but for PUFA, vitamin D and Zn( Reference Féart, Jutand and Larrieu 20 , Reference Simermann, Barberger-Gateau and Berr 21 ). Frequency of consumption of 148 food items and non-alcoholic beverages for six meals (three main, three between-meal snacks) obtained from the FFQ was transformed into discrete variables as follows: 0 for never or less than once a week, 0·25 for once a month, 0·5 for twice a month, 0·75 for three times a month, 1 for once a week and from 2 for twice a week to 7 for seven times a week. This coding was used to estimate the number of usual weekly servings of each of 148 food items recorded, ranging as a result from 0 to 49. The number of glasses of alcohol per week was also recorded. The 148 food items were then aggregated into twenty food and beverage groups (see the online Supplementary Fig. S1)( Reference Samieri, Jutand and Féart 14 ).
Dietary clusters
On the basis of the twenty food and beverage groups, we previously characterised five sex-specific dietary clusters by hybrid clustering method in 647 men and 1077 women( Reference Samieri, Jutand and Féart 14 ) (see the online Supplementary Fig. S1 for a summary of clusters identified on the 3C Bordeaux sample). In brief, the cluster analysis was based on the average number of weekly servings and performed using a mixed method combining hybrid clustering and research of stable groups during the k-means step( Reference Samieri, Jutand and Féart 14 ). As a result, among men and women, the most frequent cluster, labelled ‘small eaters’, was characterised by a lower intake of all food groups and a lower daily energy intake. The second most frequent cluster, labelled ‘healthy’, was characterised by a higher fish intake in men and a higher fruit and vegetable intake in women. In both sexes, the cluster labelled ‘Biscuits and snacking’ grouped together individuals having frequent snacks, a frequent intake of biscuits and cakes and a slightly higher energy intake. The fourth cluster was a ‘charcuterie, meat and alcohol’ cluster in men and a ‘charcuterie, starchy food and alcohol’ cluster in women. A fifth cluster included frequent ‘pasta’ eaters in men and ‘pizza, sandwich’ eaters in women.
Activity limitation
At each follow-up of the 3C study participants, mobility and limitation in IADL or ADL were investigated. Three self-reported items from the Rosow–Breslau scale were used to assess mobility: ‘doing heavy housework’, ‘walking half a mile’ and ‘climbing stairs’( Reference Rosow and Breslau 22 ). Activity limitation in IADL was assessed using the Lawton–Brody scale that assesses the self-reported ability of the participants to use a telephone, manage medication, manage money, use public or private transport and do shopping, for both sexes, and additionally to do the laundry, do housework and prepare meals for women( Reference Lawton and Brody 23 ). Limitation in ADL was assessed using five self-reported items of the Katz scale( Reference Katz, Ford and Moskowitz 24 ): bathing, dressing, toileting, transferring from bed to chair and eating. Incontinence was not considered here because it is an impairment rather than an activity limitation( Reference Spector and Fleishman 25 ). For each domain, a participant was considered disabled if he or she reported not to be able to perform at least one activity without a given level of assistance( Reference Rosow and Breslau 22 – Reference Katz, Ford and Moskowitz 24 ).
Other data
At baseline, socio-demographic information included age, sex, education (no or primary school, secondary school, high school and university), monthly income (<1500 €, 1500–2250 € and ≥2250 €) and marital status (married and divorced/separated/widowed/single). Smoking status assessed in 1999–2000 was categorised into non-smoker, ex-smoker and current smoker. BMI was computed as the weight/height squared ratio and expressed in kg/m². As fifteen out of 1328 participants had BMI<18·5 kg/m2, BMI was categorised into sex-specific tertiles. Multimorbidity was defined as ≥2 self-reported chronic diseases among cancer, hypertension, diabetes, hypercholesterolaemia, angina, cardiac rhythm disorders, cardiac failure, arteritis, myocardial infarction, asthma, Parkinson’s disease, dyspnoea, osteoporosis and thyroid diseases. Global cognitive performance was assessed using the Mini-Mental State Examination (MMSE)( Reference Folstein, Folstein and McHugh 26 ) and depressive symptomatology using the Center for Epidemiological Studies-Depression scale (CES-D)( Reference Fuhrer and Rouillon 27 , Reference Radloff 28 ). Participants underwent an extensive neurological testing for dementia, and an independent committee of neurologists made dementia diagnoses using DSM-IV criteria. Total energy intake per day was estimated from the 24-h dietary recall.
Statistical analyses
All analyses were performed separately in men and women as dietary clusters were sex specific.
We described dietary clusters based on baseline characteristics using appropriate statistics and statistical tests.
The associations between dietary clusters and the risks of mobility restriction and limitation in IADL or ADL were assessed using separated Cox proportional hazard models with age as the underlying time scale and delayed entry. The age at restriction/activity limitation onset was considered to be the age at midpoint of the interval between the diagnosis visit and the previous one as our data were interval-censored. Participants never considered as restricted/limited during the follow-up were censored at their last visit. We hypothesised that the ‘Healthy’ cluster is associated with the lowest risk of restriction/activity limitation; therefore, this cluster was chosen as the reference category. All models were adjusted for marital status, education level, income, smoking status, multimorbidity, BMI, depressive symptomatology and MMSE.
In a sensitivity analysis, we re-ran all models adjusted further for total energy intake (online Supplementary Table S1).
Statistical analyses were performed with SAS Statistical package release 9·3 (SAS Institute Inc.) and the R package ‘Survival’.
Results
Sample characteristics
Participants were aged 75·7 years on average at baseline, and they were followed up during 9·0 years on average. The 10-year incidence of mobility restriction, limitation in IADL and limitation in ADL was 12·0, 5·0 and 1·2 %, respectively, in men and 18·0, 6·7 and 1·9 %, respectively, in women.
Descriptive characteristics according to dietary clusters are displayed in Table 1 for men and Table 2 for women. Among men, 30·6 % were classified in the ‘Small eaters’ cluster, 7·7 % in the ‘Biscuits and snacking’ cluster, 25·8 % in the ‘Healthy’ cluster, 14·6 % in the ‘Charcuterie, meat and alcohol’ cluster and 21·2 % in the ‘Pasta’ cluster (Table 1). Men in the ‘Healthy’ cluster are less depressed but more likely to report multimorbidity than men in the other clusters.
Baseline socio-demographic and health characteristics of elderly men based on dietary clusters, the Bordeaux sample of the Three-City study, 2001–2002 (n 519)Footnote * (Numbers and percentages; mean values and standard deviations)
MMSE, Mini-mental state examination; CES-D, Center for Epidemiological Studies-Depression scale.
* Multimorbidity≥2 chronic diseases among cancer, hypertension, diabetes, hypercholesterolaemia, angina, cardiac rhythm disorders, cardiac failure, arteritis, myocardial infarction, asthma, Parkinson’s disease, dyspnoea, osteoporosis and thyroid diseases.
Baseline socio-demographic and health characteristics of elderly women based on dietary clusters, the Bordeaux sample of the Three-City study, 2001–2002 (n 809)Footnote * (Numbers and percentages; mean values and standard deviations)
MMSE, Mini-mental state examination; CES-D, Center for Epidemiological Studies-Depression scale.
* Multimorbidity ≥2 chronic diseases among cancer, hypertension, diabetes, hypercholesterolaemia, angina, cardiac rhythm disorders, cardiac failure, arteritis, myocardial infarction, asthma, Parkinson’s disease, dyspnoea, osteoporosis and thyroid diseases.
Among women, 30·3 % were in the ‘Small eaters’ cluster, 13·5 % in the ‘Biscuits and snacking’ cluster, 27·3 % in the ‘Healthy’ cluster, 24·7 % in the ‘Charcuterie, starchy foods’ cluster and 4·2 % in the ‘Pizza, sandwich’ cluster (Table 2).
Dietary clusters and incident activity limitation
Compared with the ‘Healthy’ cluster, men in the ‘Biscuits and snacking’ cluster were significantly at higher risk of mobility restriction (hazard ratio (HR)=3·0; 95 % CI 1·6, 5·8 adjusted for marital status, education level, income, smoking status, multimorbidity, BMI, CES-D and MMSE scores; P for global test=0·01, Table 3). Men in the ‘Biscuits and snacking’ cluster and those in the ‘Pasta’ cluster were at higher risk of limitation in IADL (HR=2·1; 95 % CI 1·1, 4·2 and HR=1·7; 95 % CI 1·0, 2·9, respectively), whereas the global test did not reach statistical significance (P=0·12). Overall, dietary clusters were not associated with the risk of limitation in ADL among men (P=0·46).
Multivariate associationsFootnote * between dietary clusters and incident functional limitations and disabilities, Three-City Study, Bordeaux, France (Hazard ratios (HR) and 95 % confidence intervals)
ADL, activities in daily living; n, analytic sample size; IADL, instrumental activities in daily living.
* Adjusted for marital status, education level, income, smoking status, multimorbidity, BMI, Center for Epidemiological Studies-Depression scale and Mini-Mental State Examination.
† Stratified on education.
‡ Stratified on smoking status.
§ Smoking status was dichotomised into never smoker v. ex-smoker and current smoker.
Among women, dietary clusters were not associated with the risk of mobility restriction and limitation in IADL. However, women in the ‘Biscuits and snacking’ cluster were significantly at higher risk of limitation in ADL than women belonging to the ‘Healthy’ cluster (HR=2·3; 95 % CI 1·3, 4·1; P for global test=0·03). The adjustment for total energy intake did not change the associations observed (online Supplementary Table S1).
Discussion
This prospective study in French older community-dwellers showed that, compared with participants belonging to the ‘Healthy’ cluster, participants in the ‘Biscuits and snacking’ cluster were at higher risk of mobility restriction and limitation in IADL in men and limitation in ADL in women over 10 years of follow-up. Moreover, men in the ‘Pasta’ cluster were at higher risk of limitation in IADL over the same period.
To our knowledge, only one study investigated the relationship between a posteriori dietary patterns and the risk of activity limitation( Reference Tomata, Watanabe and Sugawara 13 ). This 5-year study conducted in 14 260 Japanese community-dwellers aged 65 years or older identified three dietary patterns using a factor analysis: (i) the Japanese pattern characterised by high intake of fish, vegetables, mushrooms, potatoes, seaweeds, pickles, soyabeans and fruit; (ii) the animal food pattern characterised by a high intake of animal-derived food; and (iii) the high dairy pattern characterised by high consumption of dairy products, margarine and black tea. The healthier pattern,that is the Japanese one, was associated with a decreased risk of activity limitation, whereas the two other patterns were not so. Given that dietary patterns derived from no prior hypothesis are specific to population, the comparison of Japanese study’s results with our findings is difficult, especially because the cultural difference between samples is major.
The associations observed in our study are in congruence with our previous works( Reference Samieri, Jutand and Féart 14 , Reference Pilleron, Ajana and Jutand 17 ). Compared with women in the ‘healthy’ cluster, women in the ‘Biscuits and snacking’ cluster were more likely to report poor perceived health( Reference Samieri, Jutand and Féart 14 ) and were at higher risk of frailty over 10 years of follow-up( Reference Pilleron, Ajana and Jutand 17 ), whereas men in the ‘Biscuits and snacking’ cluster or the ‘Pasta’ cluster were at higher risk of muscle weakness compared with their counterparts in the ‘Healthy’ cluster( Reference Pilleron, Ajana and Jutand 17 ). Together, these data suggest that nutritional habits may interfere on the whole disablement process, from mild disorders (i.e. IADL) to severe disability (i.e. ADL).
Compared with participants in the ‘Healthy’ cluster, those in the ‘Biscuits and snacking’ cluster were characterised by higher intakes of monosaccharides and disaccharides, and low intake of fruit, vegetables and fish. Simple carbohydrates were positively correlated with inflammatory biomarkers( Reference Schulze, Hoffmann and Manson 29 ). Significant associations between inflammation and activity limitation were observed in several epidemiological studies( Reference Kuo, Al Snih and Kuo 30 , Reference Ferrucci, Harris and Guralnik 31 ). Fruit and vegetables are rich in anti-inflammatory (i.e. carotenoids) and antioxidant nutrients (i.e. vitamin C, flavonoids, polyphenols, etc.). Fruit and vegetable intake was inversely correlated with activity limitation( Reference Houston, Stevens and Cai 5 ). In a 6-year study, higher total plasma carotenoids were significantly associated with a less steep decline in 4-m walking speed and a lower risk of developing severe walking activity limitation( Reference Lauretani, Semba and Bandinelli 32 ). Low plasma carotenoid levels were associated with poor muscle strength and performance, and sarcopaenia( Reference Alipanah, Varadhan and Sun 33 – Reference Bonnefoy, Berrut and Lesourd 36 ). High oxidative stress level was an independent predictor of decline in walking speed and progression to severe walking limitation among older women followed up over 3 years( Reference Semba, Ferrucci and Sun 37 ).
Beyond probable deficiencies in micronutrients and too high intake of refined sugar, ‘Biscuits and snacking’ cluster is also characterised by destructuring of meals; this cluster may thus represent a vulnerable group regarding nutritional status but also be a marker of deterioration of the health status( Reference Samieri, Jutand and Féart 14 ).
Men in the ‘Pasta’ cluster had a higher risk for limitation in IADL. This finding is consistent with our previous ones. Men in the ‘Pasta’ cluster had poorer self-rated health and higher depressive symptoms( Reference Samieri, Jutand and Féart 14 ) that were both risk factors for activity limitation( Reference Schillerstrom, Royall and Palmer 16 , Reference Hoeymans, Feskens and Kromhout 38 , Reference Idler, Russell and Davis 39 ). The ‘Pasta’ cluster was also associated with higher risk of frailty and muscle weakness( Reference Pilleron, Ajana and Jutand 17 ). However, no mechanistic hypothesis can be formulated to explain these associations. A potential reverse causality bias may persist, although we excluded prevalent cases of activity limitation.
It was noteworthy that our results underlined a sex-specific association between dietary clusters and the risk for activity limitation. In the disablement process, functional limitation precedes activity limitation and activity limitation in IADL precedes activity limitation in ADL( Reference Verbrugge and Jette 2 , Reference Barberger-Gateau, Rainville and Letenneur 40 ). In our study, dietary clusters were associated with early stages in men (i.e. mobility restriction, limitation in IADL), whereas they were associated with the latest stage of the disablement process in women (i.e. limitation in ADL). It is widely accepted that older women suffer from higher rates of activity limitation compared with men, and some studies suggested that this fact is almost universal( Reference Wheaton and Crimmins 41 ). Indeed, women have a longer life expectancy than men; therefore, women are at a higher risk of experiencing activity limitation periods than men. The latter may die at a younger age without having experienced limitation in ADL. We cannot therefore exclude that the sex-specific association between dietary clusters and the risk of activity limitation may be due to a survival effect. In addition, the analytic sample size of men was small, which may lead to a lack of statistical power.
Our study has some limitations. Dietary clusters were built based on intake frequencies of major food groups that may not accurately reflect portions’ size and may be subject to a desirability bias. However, a previous study showed that acceptable correlations were found between number of weekly servings of foods obtained from the FFQ and the corresponding fatty acid intakes obtained from the 24-h recall( Reference Féart, Jutand and Larrieu 20 ). Another study on an independent subsample (n 105) of the Three-City Study showed also significant correlations ranging from 0·20 to 0·43 between number of weekly servings of foods obtained from the FFQ and the corresponding nutrient intakes obtained from the 24-h recall( Reference Simermann, Barberger-Gateau and Berr 21 ). Strong correlations between the two enquiries were observed for alcohol (r 0·73). However, we cannot exclude an overestimation of the validity, as participants were interviewed using the two dietary tools on the same day. Participants may have modified their dietary habits throughout the study. However, we have found a good stability in intake of fruit, vegetables, fish and meat over a 10-year period( Reference Pelletier, Barul and Féart 42 ). Information about mobility restriction, IADL and ADL was self-reported as recommended by the authors of the original scales( Reference Rosow and Breslau 22 – Reference Katz, Ford and Moskowitz 24 ). Direct observation would have probably been more accurate, but this method is very time-consuming and expensive and difficult to implement in a population-based research. However, using the self-reported tools, we could capture the participants’ feeling about their ability to do an activity that it is essential and denotes her/his engagement in the activity. Furthermore, we cannot exclude an underestimation of the incidence of activity limitation owing to a potential competing risk with mortality. Nevertheless, dietary clusters were not associated with high mortality risk in our studied sample (data not shown). Last, adjustment for potential confounding factors does not preclude for residual confounding.
The main study strengths are the 10 years of follow-up; the community-dwelling population-based design; the innovative approach with a mixed clustering strategy to identify dietary clusters; and comprehensive adjustment for several confounders.
Conclusion
In this prospective study in French older community-dwellers, some sex-specific dietary patterns seem to be deleterious for activity limitation in the long term. Improving and maintaining healthy dietary habits in older adults may be beneficial for maintaining autonomy and delaying the onset of activity limitation and thus increasing activity limitation-free life expectancy.
Acknowledgements
The Three-City Study is conducted under a partnership agreement between the Institut National de la Santé et de la Recherche Médicale (INSERM), Victor Segalen – Bordeaux 2 University and the Sanofi-Synthélabo company. The Fondation pour la Recherche Médicale funded the preparation and beginning of the study. The 3C-Study is also sponsored by the Caisse Nationale Maladie des Travailleurs Salariés, Direction Générale de la Santé, Conseils Régionaux of Aquitaine and Bourgogne, Fondation de France, Ministry of Research-INSERM Program Cohortes et collections de données biologiques, the Fondation Plan Alzheimer (FCS 2009-2012) and the Caisse Nationale pour la Solidarité et l’Autonomie (CNSA). The authors finally thank Mélanie Le Goff for her help on statistical analysis.
This work was supported by the European Union’s Seventh Framework Programme (FP7/2007-2013) FRAILOMIC Project (grant no. 305483) to S. P. The study sponsors played no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; and preparation, review or approval of the manuscript.
The authors’ contributions are as follows: S. P. formulated the question, conducted the data analysis and wrote the first draft; C. F. was involved in the data analysis and interpretation of the data. All authors reviewed the manuscript, provided further contributions and suggestions and approved the final draft.
C. F. received fees for conferences from Danone Research and Nutricia. The other authors declare no conflicts of interest.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114518001654
